Potential Savings from Optimized Plan Choice in Medicare Beneficiaries with Schizophrenia Abstract The Medicare prescription drug benefit (Part D) relies on private market mechanisms to deliver care: a typical beneficiary can choose from between 45 and 57 private stand-alone Part D plans (PDP) in 2009, depending on his/her region of residence. It is known that beneficiaries do not choose plans based on their medication needs; and that the ability to choose plans is poorer among beneficiaries with mental disorders. The purpose of this study is to simulate personalized plan choice based on patient medication needs in schizophrenia and quantify potential improvement in patient drug coverage and potential savings to the government. Part D provides substantial premium and cost-sharing assistance to beneficiaries qualifying for the low-income subsidy (LIS) program. About 93% of PDP enrollees with schizophrenia received the LIS in 2007, whereas 40% of all PDP enrollees did. The majority of LIS enrollees are randomly assigned to PDP plans with premiums at or below the regional average. Random assignment does not assign enrollees to plans based on their medication needs and has caused severe problems including disruptions in plan coverage for 5.9 million between 2007 and 2010; and high beneficiary and government spending for those assigned to plans requiring high copayments. Our study will develop intelligent assignment algorithms based on beneficiaries' medication needs and dynamics of plan features in the Part D market. The algorithms can be easily implemented each year after an initial setup of software without substantial costs. The intelligent assignment method can be used by the government to assign/reassign beneficiaries or provide personalized assistance to help beneficiaries with schizophrenia to choose plans. The intelligent assignment method has the potential to substantially reduce government spending while improving patient outcomes.